Just so that you don't dismiss me as merely "whining" about how bad this most recent roll-out was - I'm just gonna use YOUR tutorial to demonstrate it. You're not creating a tool for artists - you're adding a toy for people who just want to swap in quick backgrounds for play, which would be fine - except you are doing a very poor job at even that. Additionally, artists are paying good money and you keep making promises that the slow-downs, breaks, constant updates, and problems are because you're expanding what's available to us (to either justify the cost, finally, or raise it on us... who knows at this point). However, look at bgissue2.jpg. I started YOUR tuorial, let it load the picture, and then followed the instructions: Remove background. And, it did. Photoshop AI chose the subject (good luck choosing the subject on ANYTHING with ANY complexity - that's totally useless, I've tried on 2 dozen other pictures with terrible outcomes), and it cut off part of the bottle in YOUR tutorial. That's funny. So, look at bgissue1.jpg (sorry about the poor quality of the images / fuzziness, but resizing in photoshop is really, really bad still. I mean - you could've spent all this time using ai to treat rasterized images like vectors and allow for resizing, but... you didn't, so the quality just continues to be terrible and it makes your whole "combining photos" thing you keep trying to push... not really feasible. But, that's also just more evidence of how bad the company's prioritization is). Sorry, getting back to bgissue1.jpg, while it's understandable that somehow AI can't create the same image twice to look like your tutorial (or... it's supposed to be understandable?), unlike your video prompt, it created shadows to the shadows of the bottle. That sucked. On the top picture, it created the background as two pieces of paper and added shadow to the line? It didn't even bother on the bottom photo to add shadow around the flowers on the right. (Oh, and I had to widen the canvas manually because the picture you include does not match the picture in the video, but that's okay, if you didn't care enough to get the software right and you're cheating the prompting, there's no real point in taking the time to get the full instructions right). Now we look at bgissue3.jpg. So, your prompting in the video was clearly, not honest by the comparitive outcome being so vastly different from the generated pictures to the video output. Hey, maybe misleading people to get subscriptions pays off, I don't know? Of course, since AI prompting is something I have to do for part of my living, bgissue3.jpg will show you the outcome of my work, much closer to the tutorial you provided and some of the extended prompting I had to apply. I didn't refine it further because there was no need, I demonstrated what I was claiming above about this continued "ai" roll-out being problematic and poor quality. Those of us who do AI prompting are accustomed to the requirement of added descriptions and some specificity (and about 500 iterations), so it's not a big deal. But, from your photoshop AI "subject selection" cutting off your own bottle, to the prompting not matching the output and the horrible errors occurring repeatedly, your "generate background" feature, just like the others I critiqued above, is seriously lacking in functionality and you're just pushing for more and more subscribers without being completely forthcoming about your AI: it's restrictions, limitations, or functionality. So, all these other fake "bot" responses in here, "oh, this was so easy and so much fun and wonderful" .... are SO bad... no one could EVEN follow your tutorials and get anything of quality or close. I don't know if you're using AI to generate the fake responses - but this is truly the lowest end treatment of customers from a billion dollar corporation making hundreds of millions per month on the profit side. Good luck with your next update! (final note-  you'll see that I circled the right hand side of the sprayer in the top picture of bgissue1.jpg... that's because the flowers would generate a reflection. We could easily add that ourselves, but it's one more weakness of the system you've created.)